Diablo® III

CM Proc Rate (Testers needed)

UPDATE: TekkZero did some testing on sig spells and the results indicate that CM is working as we previously expected for a few runes. WW still is not working as expected. Channelled spells like RoF and Disintegrate are still in need of testing and confirmation for sig spells is needed.

WW testing will need to include APS and CC information. See http://us.battle.net/d3/en/forum/topic/7788988884?page=2#31 for preferred methodology, but feel free to post your own results with how you obtained them.

Original Post:
Recent results from a couple people have suggested that the CM proc rate might not be as straight forward as we assume. Currently, the general consensus is that CM reduces the cooldown of all your spells after you crit and pass a check against the spell coefficient. For WW that means 12.5% of your WW tics that crit would proc CM. There is now cause to think that CM works in a slightly different manner.

The first post I saw with some evidence contrary is here by UBORLOL http://us.battle.net/d3/en/forum/topic/6794871641?page=25#481

Steve's DPS simulator already uses a CM coeffcieient of 2 because he noticed his in game results weren't lining up properly with the simulator results. Meaning for WW 25% of your WW tics would proc CM instead of 12.5%. He has some data findings on his simulator page, but I haven't had a chance to look through those on account of being an old fashioned IE user. His simulator thread is linked below, and links to his simulator can be found there.
http://us.battle.net/d3/en/forum/topic/7709121181

I have also done some limited testing and my one test case gave me 37 DS casts over a 88s fight against Ghom where I did nothing except cast DS and WW. DS rune was crystal shell so my only CM procs should have been from WW and white attacks, and I only had a couple white attacks. Simulating the results with our original assumption of how CM works shows I should only average 24 DS casts in that time frame, and after 500 simulations the most casts I should have seen was 27, a full 10 less than I actually got. In other words, I'm seeing a lot more CM procs in game than I should be.

These results are actually consistant with my SA examination in which I found SA was procing quite a bit more often from WW as it should.

What I'd like to do is ask for some people to do a few tests on their own. The idea is record a fight, preferably 40-60s+, then count how many DS casts you get off. For the fight just use 1 spell and DS. I would love to have data from several different spells to see if the behavior is specific to dot type spells like WW or also applies to sig spells and such.

List of Spells:
WW
MM
Disintegrate
RoF
Shock Pulse
Spectral Blades
Electrocute
Blizzard

For those willing to help test, pick a spell and rune and record DS casts and fight duration, as well as attack speed and crit chance. Your dps doesn't matter and 1-2 fights should be fine for the test, so don't worry about carrying out 20 tests to average over. Preferably do the fights against Ghom, and run to a corner at the start of the fight, wait for him to reach you, then spam the spell of choice while refreshing DS as often as possible.
Edited by Loroese#1415 on 1/29/2013 12:58 PM PST
Reply Quote
Nice, I think it would be better to stress doing any test in triplicate as this provides far more robust data without requiring too much effort, say, as a 20 repeat test would

1 test could be marred by some unknown error or something, so having another 2 that line up consistently is a good way of identifying this. With just 1 test you never know that your 1 execution of it was anomalous

I would love to partake in this type of testing but
A) computer = mac laptop, average around 1k ping
B) can't record anything while playing, would cause even more lag
Reply Quote
This is a double-post from the WW Tick Rate and Breakpoint Thread. Here's my tests:

TESTS: Hit my brother's barb in PVP (1.07 PTR) for 6 total refreshing of diamond skin (crystal shell). I tried twice for each skill setup:
magic missle (seeker) --> 6 refreshes took 55s and 53s
shock pulse (piercing orb) --> 6 refreshes took 70s and 71s
magic missle (penetrating blast) --> 6 refreshes took 78s and 77s

EDIT: I thought i had the math predicting the cooldown, but i didn't take into account a few factors.
Edited by TekkZero#1963 on 1/28/2013 6:40 PM PST
Reply Quote
EDIT: data already posted

Going to look over the math real fast.
Edited by Loroese#1415 on 1/28/2013 6:18 PM PST
Reply Quote
I made a mistake on the bottom calculations. It's not 1.5x. I'm right now figuring out what best fits the data.
Reply Quote
To be clear about how you did the testing, was it cast DS, then spam spell until you could cast 6 MORE DS? So with no spell casting it'd take 12.75s to cast DS1, then another 12.75 to cast DS2, up to 76.5s to cast DS6? Or was there another method?
Reply Quote
yeah that's correct. i just kept spamming the primary spell and kept trying to refresh diamond skin. i counted until i had refreshed six times. So i think what i failed to do in my above analysis is take into account that there's a natural cooldown. im gonna re-crunch my numbers taking that into account.
Edited by TekkZero#1963 on 1/28/2013 6:34 PM PST
Reply Quote
Here's the model I developed to compare:

Assuming you use Evocation, the CD on DS is 15*0.85 = 12.75s.

Expected CM procs per second (pps) = APS*CC*Spell_Coef
Expected time between DS casts = 12.75 / (1+Expected CM pps)
It's 1+Expected procs because each second of casting spells also removes 1s from the CD.

If I plug in the numbers for the 3 spells above, we get 7.12, 9.14, 10.09s between casts for 1.58 APS and 50% crit for spell coefficients of 1, 0.5, 0.3333 respectively.

Expected time for 6 casts with no modified CM coefficient:
Seeker: 42.74s
Piercing Orb: 54.84s
Penetrating Blast: 60.55s

So the results don't fit either model at all. If the old model were wrong, I'd expect times below those shown above.
Edited by Loroese#1415 on 1/28/2013 6:43 PM PST
Reply Quote
One possibility is also a bit of dodge. My bro's barb had 14.3%. I'll try these tests again later. maybe we can build up a library of more datapoints to measure against.
Reply Quote
Ah, dodge. I guess that's a good reason to test against Ghom or Azmo.
Reply Quote
I can't help but fear this will lead to a nerf of some sort in the future.
Reply Quote
this topic should be redacted/deleted. it can only lead to bad things
Reply Quote
Hey Loroese could you edit your post in the WW breakpoint thread that quotes my initial results? I'd like to eliminate that initial analysis i did that didn't take into account the natural cooldown. I don't want someone reading that and getting confused. The original data is "fine" as far as how I collected it. Though i agree we should re-do the tests probably on ghom since dodge may complicate matters.

@atinybug, chrisloup: I'm not sure I really buying into the argument. You could make the argument that us investigating how WW scales with attack speed could have ended in a nerf. I'd wager I think if there is a multiplier, it's probably intended. Blizzard has never mentioned that all abilities follow the LoH proc coefficients exactly. This could be a situation where we just misunderstood the mechanics of how actually CM works and that there was always a multiplier.
Edited by TekkZero#1963 on 1/28/2013 8:19 PM PST
Reply Quote
Sure can do.

Also, I agree that this is likely an intended effect that Blizz is aware of since they actually wrote the code interactions of CM and all the spells. I think it's just that we assumed CM worked a certain way, because it made sense based on other interactions. Now, it seems those assumptions weren't completely right. It just remains to be seen if the interaction is dot specific or applies to all skills, thus my desire for the test cases.

If it's not intended, I still don't think that matters much because Blizzard is fully aware of how CMWW is working and what it can do. If they nerf the build they nerf it because of how it works, not because we suddenly find a bug in the CM proc rate. They've already been lowering the spell coefficients to reduce the CM procs anyway. If they want them lower, they'll do something to lower them again.
Edited by Loroese#1415 on 1/28/2013 8:22 PM PST
Reply Quote
Ghom Test - MP2 (No Evocation) - Latency (120-150ms)
Attack Speed = 2.59aps
Crit Chance = 55.5% cc
Expected Cooldown Period for Diamond Skin = 15s*6 = 90s

TEST DURATIONS:
*Description: Just held-down my left click for the primary and was constantly tapping Diamond Skin.
Seeker with 6 Diamond Skin Refreshes: 42s and 43s
Penetrating Blast with 6 Diamond Skin Refreshes: 67s and 68s
Shock Pulse with 6 Diamond Skin Refreshes: 55s and 56s

Number Of Attacks with Seeker = 2.59aps*42.5s = ~110
Number Of Attacks with Penetrating Blast = 2.59aps*67.5s = ~175
Number Of Attacks with Shock Pulse = 2.59aps*55.5s = ~144

Proc Coefficient = (Expected Cooldown Period - Test Duration) / (Crit Chance*Number Of Attacks)
Seeker: 0.7775 (Known Coefficient: 1.00)
Penetrating Blast: 0.23 (Known Coefficient: 0.33)
Shock Pulse: 0.43 (Known Coefficient: 0.5)
Average Scaling Factor: 0.778x (presumed CM "multiplier")

So I just did Ghom and if I did the test right, it actually makes things even less definitive. It appears to be a 0.75x multiplier (ie it's less efficient than LoH). Can this actually be right or am I missing something in the math? I tried Wicked Wind too and the measured "CM multiplier" was even worse. But I can't rule out that I'm doing a horrible job of centering my twisters on Ghom. For the WW tests, I moved up into the upper left corner and waited over 6s, so I would be at full-wind up before I started the test.
Edited by TekkZero#1963 on 1/28/2013 10:50 PM PST
Reply Quote
Do you guys think this could be an RNG 'clumping' issue? The RAND generator being used to decide if each crit 'passes' the coefficient for a CM proc is having some machine memory issues? It seems we're getting alot of odd results that arn't following any real pattern (some below expected, some above, and most difficult to repeat)

I got a couple days off work this week with nothing going on as of yet. I'll put some time into this and see whats up. We only tested coefficients through use of LoH and then made the assumption all the spells that said 'chance of XXX occuring 'on hit (crit)'' used the exact same proc coefficient.

Its entirely possible (although I find it somewhat unlikely) that some of these skills have a 'coefficient of a coefficient' attached to them. Loroese's work on SA however seems to support that there are too many bolts (alot too many). If there is one exception there can be more.
Reply Quote
Those results seem good, and I get the same Proc Coefficient with your numbers. I think the difference in the results is just a matter of latency and varience. My take home results from this are that sig spells seem to behave as we previously thought with respect to CM procs, but I'd still like to see some results from other spells or other people to confirm the behavior.

What sort of results are you seeing from WW test? All I do is sit at the upper left corner until Ghom waddles into melee range, then I point at him and hold my WW button. I don't bother with trying to place the WWs since he shouldn't really move out of them.
Reply Quote
Do you guys think this could be an RNG 'clumping' issue? The RAND generator being used to decide if each crit 'passes' the coefficient for a CM proc is having some machine memory issues? It seems we're getting alot of odd results that arn't following any real pattern (some below expected, some above, and most difficult to repeat)

I got a couple days off work this week with nothing going on as of yet. I'll put some time into this and see whats up. We only tested coefficients through use of LoH and then made the assumption all the spells that said 'chance of XXX occuring 'on hit (crit)'' used the exact same proc coefficient.

Its entirely possible (although I find it somewhat unlikely) that some of these skills have a 'coefficient of a coefficient' attached to them. Loroese's work on SA however seems to support that there are too many bolts (alot too many). If there is one exception there can be more.


I'm actually suspecting it has something to do with WW itself and the dot mechanic. I saw similar behavior back when I was looking at SA, like you mentioned, and noticed that the 1.0 coefficient sig spells was proccing SA everytime it crit for exactly 35% weapon damage, so there was nothing funny going on there. Then when I used WW it seemed to be doing 2.5x the expected dps, give or take depending on APS. Here, we're seeing something like 1.5-2x the expected CM procs, but so far only for WW, so maybe we don't quite understand WW as well as we think. The part that trips me up is that we have the LoH mechanics figured out very well, so I don't see why the crit mechanics would be different.

If I had the time I'd test to try and find the WW CM rate at various APS breakpoints to see if those breakpoints have any impact on it, in addition to testing a few other spells. The problem is I have neither the time nor the gear for higher APS breakpoints.
Reply Quote
@Loresse: Good to know you get the same results on the primaries.

After muling over it, I realized what I had done wrong in my initial calculations for WW and why I was getting such weird numbers (ie for the calculation of the number of hits, I used the Tick Rate and not the Tick Rate per second). So here they are now that I'm confident I'm getting reasonable numbers. Moreover, I'm getting the 2x CM Multiplier we've been suspecting existed. So it sounds like Steve's assumption on his CM/WW wiz simulator is right on.

Here's the tests for WW:
Expected Cooldown Period for Diamond Skin = 15s*16 = 240s
2.59aps --> 21s and 22s
1.88aps --> 30s and 31s

Number Of Hits with WW (2.59aps) = (2.59aps*6)*(33/6)*21.5s = ~1,838
Number Of Hits with WW (1.88aps) = (1.88aps*6)*(24/6)*30.5s = ~1,376

Proc Coefficient = (Expected Cooldown Period - Test Duration) / (Crit Chance*Number Of Hits)
WW (2.59aps) = 0.2142 (Known Proc Coefficient: 0.125)
WW (1.88aps) = 0.2743 (Known Proc Coefficient: 0.125)
Mean WW Proc Rate = 0.244
Average Scaling Factor: 1.95x (presumed CM "multiplier")

So now I wonder how other DoTs (like WW) work with CM and whether there's a systematic difference from skills like primaries. If I have time tomorrow, I'd like to try some tests for Ray of Frost (Cold Blood) and Meteor (Star Pact); however, I'm unsure on the exact math I should be using to predict the number of hits. My understanding is that it's suppose to be a constant 1s tick rate for Meteor, but what about Ray of Frost?

Regardless, it does now sound like both you and Shandlar's intuition might be correct that are coefficients within coefficients. It'll be interesting to see if it's unique to DoTs or is skill-specific.
Edited by TekkZero#1963 on 1/29/2013 12:09 AM PST
Reply Quote
I find it interesting that you see such a difference between the attack speeds (though that might have some to do with AP usage). I wouldn't average the value when the test cases are different like that.

You are now the 3rd person I have seen with data suggesting a higher CM proc rate for WW than expected, so I think we're definitely onto something. I'd really like to see similar analysis for Disintegrate and RoF (maybe with Prism rune on DS to make it easier to manage AP), as well as a couple other results using spammed primaries.
Reply Quote

Please report any Code of Conduct violations, including:

Threats of violence. We take these seriously and will alert the proper authorities.

Posts containing personal information about other players. This includes physical addresses, e-mail addresses, phone numbers, and inappropriate photos and/or videos.

Harassing or discriminatory language. This will not be tolerated.

Forums Code of Conduct

Report Post # written by

Reason
Explain (256 characters max)
Submit Cancel

Reported!

[Close]